T.N. DAVIES and D.D. HOFFMAN
FACIAL ATTENTION AND SPACETIME FRAGMENTS
ABSTRACT. Inverting a face impairs perception of its features and recognition of its identity. Whether faces are special in this regard is a current topic of research and debate. Kanizsa studied the role of facial features and environmental context in perceiving the emotion and identity of upright and inverted faces. He found that observers are biased to interpret faces in a retinal coordinate frame, and that this bias is readily overruled by increased realism of facial features, but not easily overruled by environmental context. An additional factor contributing to a retinal coordinate-frame interpretation may be the ambiguous nature of the face stimuli. Since his facial expressions are interpretable both upright and inverted, they may in both orientations activate an endogenous attentional pro- cess for faces. We present visual search and change-blindness experiments that explore how inversion, negation, and facial emotion affect visual attention to static faces. We find that attention to faces is impaired by inversion and negation. We also find that the parts of the face that receive greater attention can be influenced by the emotional expression of the face. We propose to extend these experiments to dynamic faces. To this end, we develop a theory of the visual representation of dynamic faces, in which faces are represented by classes of ‘spacetime fragments’-moving regions of the face with high informational content. We then present ideas for future experiments which are motivated by the spacetime fragment theory, and which should serve to constrain its further development.
1. KANIZSA’S STUDY OF FACES
Among the many topics to which Kanizsa applied his talents as an artist and experimental psychologist was the perception of faces (Kanizsa and Tampieri 1976). He was interested in whether the perceptual orientation of visual objects depends on a retinal or an environmental frame of reference. Some research had suggested the importance of the environmental frame (James 1890; Miyakawa 1950), some the importance of the retinal frame (von Helmholtz 1867; Thouless 1947), and some the importance of viewer expectations (Rock 1956, 1973). Kanizsa noted that these differing results were obtained with different kinds of visual objects, and proposed that study of the specific properties of these visual objects could explain the different results. To this end he